Horse tissues were submitted as formalin-fixed excisional biopsies to the Alfort Veterinary Medicine School (n = 12), and at the IDEXX Diagnosis Laboratory (n = 27). Skin samples consisted in previously diagnosed cutaneous melanomas (n = 9) or melanocytomas (n = 15) (Table 1), and in normal glabrous skin (n = 6) from under the tail area and lips, as well as normal furry skin from the rump (n = 6), used as control. Tissues that were not archival diagnostic material were taken at the programmed euthanasia of individuals used for another study. Euthanasia had been practiced following recommended consensus publications. Tumors were selected such that cytological characteristics or vascular emboli indicated a clear diagnosis. Briefly, dermal or epidermal proliferations of nests of melanocytes with variable degrees of pigmentation were considered melanocytomas when they were well delimited, without or with low atypia, no mitoses or less than 2 mitoses per ten high power fields, and no signs of vascular invasion or vascular embolus. Melanomas were characterized by anisocytosis, anisokaryosis, prominent nucleoli, nuclear atypia or alternatively tumoral vascular emboli. A high mitotic index corresponding to more than 10 mitoses per ten high power fields was also considered a mark of malignancy, however samples with low or moderate mitotic index could carry other characteristics of malignancy.

M: Male; F: Female, G: Gelded, -: unavailable data.

Age of animals, when available, ranged from 2 to 25 years. In our sampling aged gray horses have a low representation (Table 1) despite the high incidence of slow evolving melanocytic tumors in these horses 16 17 .

The protocol for immunofluorescence was as previously described 8 with minor modifications. Briefly, antigen retrieval was performed in Tris-EDTA, pH 9, for 30 min in a water-bath. Antibodies were mouse monoclonals anti-MITF (Zymed, dilution 1:50; Invitrogen, Cergy-Pontoise, France) anti-RACK1 (Transduction Laboratories, 1:150; BD Biosciences, Le Pont de Claix, France) and rabbit polyclonals anti- cytokeratin5 (Covance; 1:1000; Eurogentec, Angers, France) and anti-PAX3 (Zymed; 1:200). Nuclear counter-staining was achieved with 4', 6'-diamidino-2-phénylindole (DAPI) (Invitrogen, 1:1000). Sections were examined with a Zeiss Axio Observer Z1M ApoTome microscope (Carl Zeiss S.A.S. ; Le Pecq, France). Controls without the first antibodies showed no unspecific labeling. Images were processed with the AxioVision computer program version 4.6 (Carl Zeiss). Figures are representative of the skin samples evaluated. All images shown are individual sections of z series stack. Final figures were assembled with Adobe Photoshop CS3 (Adobe Systems; USA).

RACK1 staining distribution was analyzed at the tissular and cellular levels. Distribution within the cytoplasm was graded 0 when homogeneous and 1 when heterogeneous. Samples were graded blindly without reference to pathology reports.

All histopathological evaluations were carried out on routinely stained hematoxylin-eosin-safran sections. The size of the tumors was missing in some clinical files; the dimensions of the lesions on histological sections ranged from 4 to 80 mm. To highlight the histological specificity of horse melanomas and melanocytomas, all tissue sections were examined at 3 different magnifications (10, 20, 40 high power field) and classified according to the eleven histo-morphometric criteria previously defined by Viros et al. 4 : scatter of intraepidermal melanocytes, nest formation of intraepidermal melanocytes, cytoplasmic pigmentation of neoplastic melanocytes, size and shape of cells, nuclei and nucleoli, epidermal contour, lateral circumscription, thickness of normal epidermis and presence of ulceration. Grading was carried out like in Viros et al. 4 except for ulceration which was graded 0 for absence and 1 for presence. The samples were graded blindly independently by two of us without reference to pathology reports, Two groups –melanomas and melanocytomas– were subsequently made based on these reports. Sections were observed with a Leica DMLB microscope (Leica Microsystems S.A.S., Nanterre, France). Images were processed with the MetaVue Imaging System (Molecular Devices; St Grégoire, France) computer program. Histological pictures were taken with AxioImager.ZI through a AxioCam HRc camera and processed with the AxioVision 4.6.3 SPI software (Carl Zeiss).

Statistical differences between means taken in pairs were evaluated by Student’s t test. The test was adapted for a number of samples below 30 18 . A P-value <0.05 was considered as statistically significant.

In order to analyze melanocytic proliferations, we first needed a marker to identify melanocytic cells within tissue sections. Both MITF and PAX3 transcription factors are expressed by melanocytes and their precursors 19 20 . Comparison of horse and human protein sequences for MITF and PAX3 resulted in more than 90% identity. On tissue sections, melanocytes at the basal layer of healthy skin were labeled by a specific nuclear signal using a rabbit PAX3 antibody (Figure 1 A1) or a mouse MITF antibody (Figure 1 C1). Unspecific labeling was not detected (not shown). Moreover, MITF and PAX3 are expressed by melanocytic cells within tumoral proliferations 8 21 . PAX3 and MITF-positive cells were identified both in melanocytomas and in melanomas regardless of the pigmentation of the lesion (Figure 1 A2, A3, C2, C3). Both MITF and PAX3 antibodies proved to be helpful in identifying the melanocytic lineage in horse tissues. Nevertheless, MITF antibody detected the nucleus of melanocytic cells with more sensitivity than did the PAX3 antibody. MITF antibody was thus used for further analyses.

We analyzed the tissue and cellular distribution of RACK1 protein, in healthy skin and melanocytic lesions, after double immunostaining for RACK1 and MITF. In healthy control skins (n = 12), RACK1 protein was highly expressed in the cytoplasm of keratinocytes which were used as positive controls (Figure 2A). By contrast, MITF-positive melanocytes were negative for RACK1 (Figure 2A). Triple immunostaining against RACK1, MITF and cytokeratin 5 (CK5), a marker of basal keratinocytes, was performed in order to better identify the cell type containing RACK1 in the epidermis. Every signal for RACK1 in the vicinity of melanocytes colocalized with CK5 (Figure 2B). Thus, in these labeling conditions, RACK1 was not detected in horse normal melanocytes.

In all melanocytic lesions examined (n = 24), RACK1 was extensively detected in MITF-positive cells, with two distinct distribution patterns. RACK1 was distributed either heterogeneously (Figure 3A) or homogeneously over the lesion, whether tumors were pigmented or not (Figure 3B, 3C). In lesions with heterogeneous distribution, RACK1 was detected as a granular cytoplasmic staining in melanocytic cells. In sharp contrast, in lesions with homogeneous distribution, RACK1 staining was diffuse, perinuclear and cytoplasmic (compare Figure 3A with Figure 3B and 3C). We tested whether the cytoplasmic distribution of RACK1 staining could be of help in classifying melanoma lesions. For this purpose the distribution of RACK1 was graded 0 when homogenous, and 1 when heterogeneous. Samples were graded independently by two of us, blindly without reference to pathology reports. Subsequently, melanocytoma and melanoma samples were grouped based on pathology reports. Comparison of RACK1 grading between the two groups resulted in a statistical difference (1 ± 0 vs. 0.2 ± 0.36 respectively; P < 0.001). All samples histologically classified as melanocytomas (n = 15) stained heterogeneously for RACK1. Among melanomas (n = 9), two samples also stained heterogeneously for RACK1 (data not shown). Noteworthy, these two melanomas had several characteristics of low histopathological aggressiveness, which was low mitotic rate and no vascular embolisation but high anisokaryosis. All other melanoma samples stained homogeneously for RACK1. Thus, cytoplasmic RACK1 labeling may be helpful in distinguishing melanoma from benign melanocytic skin tumors.

Melanocytic lesions in our samples displayed a large histological variety. Figures 4A and 4B show various pigmentation patterns and cell shapes in two representative melanomas. Interestingly enough, RACK1 distribution pattern was constant throughout the whole of the lesions, as illustrated in Figure 4C. To better characterize the lesions, we performed a detailed analysis based on the morphological criteria Bastian’s group used on human samples 4 , summarized in Table 2. We found that equine melanomas were characterized by: high pigmentation (Table 2, 2^nd column), abrupt lateral circumscription (Table 2, 4^th column) at the transition from involved area to adjacent normal tissue (Figure 4A) and a thicker epidermal contour (Table 2, 5^th column), indicative of epidermal hyperplasia. Moreover, we often observed an absence of junctional component, as previously described 17 . However none of these characteristics reached statistical significance when compared to those of melanocytomas. Finally, we checked every morphological criterion and the RACK1 distribution pattern against the malignancy status. RACK1 signal pattern of distribution was more frequently indicative of malignancy than any single morphological criterion, pointing out its usefulness as a diagnostic marker.

^aScatter of intraepidermal melanocytes: 0, absent; 1, slight; 2, medium; 3, prominent.

^bCytoplasmic pigmentation of neoplastic melanocytes: 0,absent; 1, slight; 2, medium; 3, high; 4, very high.

^cNesting of intraepidermal melanocytes: 0, absent; 1, slight; 2, medium; 3, prominent.

^dLateral circumscription: 0, discontinuous; 1, continuous; 2, abrupt.

^eEpidermal contour: 0, atrophy; 1, effacement; 2, normal; 3, thickening; 4, hyperplasia.

^fCell size: 1, small; 2, medium; 3, large.

^gCell shape: 0, round; 1, ovoid; 2, elongated; 3, spindled.

^hNuclear size: 1, small; 2, medium; 3, large.

ⁱNuclear shape: 0, round; 1, ovoid; 2, elongated; 3, spindled.

^jNucleolar size: 1, small; 2, medium; 3, large.

^kUlceration: 0, absent; 1, present.

^lRACK1 distribution pattern: 0, homogenous; 1, heterogenous.

-: Impossible evaluation.

x.5: means of two different authors observations.